decommissioning datanodes
Hello, I'm in the trying to figure out how to decommission data nodes. Here's what I do: In hdfs-site.xml I have: property namedfs.hosts.exclude/name value/opt/hadoop/hadoop-1.0.0/conf/exclude/value /property Add to exclude file: host1 host2 Then I run 'hadoop dfsadmin -refreshNodes'. On the web interface the two nodes now appear in both the 'Live Nodes' and 'Dead Nodes' (but there's nothing in the Decommissioning Nodes list). If I look at the datanode logs running on host1 or host2, I still see blocks being copied in and it does not appear that any additional replication was happening. What am I missing during the decommission process? -Chris
Re: decommissioning datanodes
Do you mean the file specified by the 'dfs.hosts' parameter? That is not currently set in my configuration (the hosts are only specified in the slaves file). -Chris On Fri, Jun 8, 2012 at 11:56 AM, Serge Blazhiyevskyy serge.blazhiyevs...@nice.com wrote: Your nodes need to be in include and exclude file in the same time Do you use both files? On 6/8/12 11:46 AM, Chris Grier gr...@imchris.org wrote: Hello, I'm in the trying to figure out how to decommission data nodes. Here's what I do: In hdfs-site.xml I have: property namedfs.hosts.exclude/name value/opt/hadoop/hadoop-1.0.0/conf/exclude/value /property Add to exclude file: host1 host2 Then I run 'hadoop dfsadmin -refreshNodes'. On the web interface the two nodes now appear in both the 'Live Nodes' and 'Dead Nodes' (but there's nothing in the Decommissioning Nodes list). If I look at the datanode logs running on host1 or host2, I still see blocks being copied in and it does not appear that any additional replication was happening. What am I missing during the decommission process? -Chris
Re: decommissioning datanodes
Thanks, this seems to work now. Note that the parameter is 'dfs.hosts' instead of 'dfs.hosts.include'. (Also, the normal caveats like hostnames are case sensitive). -Chris On Fri, Jun 8, 2012 at 12:19 PM, Serge Blazhiyevskyy serge.blazhiyevs...@nice.com wrote: Your config should be something like this: property namedfs.hosts.exclude/name value/opt/hadoop/hadoop-1.0.0/conf/exclude/value /property property namedfs.hosts.include/name value/opt/hadoop/hadoop-1.0.0/conf/include/value /property Add to exclude file: host1 host2 Add to include file host1 host2 Plus the rest of the nodes On 6/8/12 12:15 PM, Chris Grier gr...@imchris.org wrote: Do you mean the file specified by the 'dfs.hosts' parameter? That is not currently set in my configuration (the hosts are only specified in the slaves file). -Chris On Fri, Jun 8, 2012 at 11:56 AM, Serge Blazhiyevskyy serge.blazhiyevs...@nice.com wrote: Your nodes need to be in include and exclude file in the same time Do you use both files? On 6/8/12 11:46 AM, Chris Grier gr...@imchris.org wrote: Hello, I'm in the trying to figure out how to decommission data nodes. Here's what I do: In hdfs-site.xml I have: property namedfs.hosts.exclude/name value/opt/hadoop/hadoop-1.0.0/conf/exclude/value /property Add to exclude file: host1 host2 Then I run 'hadoop dfsadmin -refreshNodes'. On the web interface the two nodes now appear in both the 'Live Nodes' and 'Dead Nodes' (but there's nothing in the Decommissioning Nodes list). If I look at the datanode logs running on host1 or host2, I still see blocks being copied in and it does not appear that any additional replication was happening. What am I missing during the decommission process? -Chris